TW202201918A - System for synthesizing signal of user equipment and method thereof - Google Patents

Abstract

A system for synthesizing signal of user equipment and a method thereof are provided. The system includes a physical channel modeler and a physical channel training module. The physical channel modeler receive a geo information of a field under test of and a sparse real physical field channel feature to build a physical channel model. The physical channel modeler estimate a plurality of predefined positions of the geo information to obtain a plurality of simulated physical field channel features corresponding to the predefined positions. The physical channel training module receives and performs training on the geo information, the sparse real physical field channel feature and the simulated physical field channel features by using an AI algorithm to perform an inference of a fully real physical field channel feature.

Description

Virtual user device signal synthesis system and method

The present invention relates to a signal synthesizing system and method thereof, and more particularly, to a signal synthesizing system and method for virtual user equipment.

Generally speaking, when the base station equipment manufacturer needs to perform the function and performance of the base station performance test for each geographic location in a physical field to be tested, there are two existing methods. One method is to use a large number of terminal devices or a single device to continuously perform point-by-point measurements to obtain complete real entity field channel characteristics. The former causes a lot of waste of measuring equipment and manpower, while the latter causes waste of measurement time and manpower. In addition, the measurement results may also be outdated or inaccurate due to changes in facilities in the physical site, climate, temperature and other environmental factors. As a result, it will result in a situation where the measurement must be re-measured and cannot be estimated or updated.

Another method is to use the entity layer channel modeler to model the channel according to the entity to be tested, so as to obtain the channel characteristics of the simulated entity domain. However, this approach models the channel assuming ideal conditions. Under the premise of excluding all imperfect phenomena, the simulation results may also be greatly inaccurate.

The present invention provides a virtual user device signal synthesis system and the system thereof, which can provide fast and high-accuracy simulation results.

The present invention provides a virtual user device signal synthesis system, which includes a physical channel modeler and a physical channel training module. The physical layer channel modeler receives a geographic information of a test field and a sparse real physical field channel feature (Sparse Real Physical Field Channel Feature) to build a physical layer channel model. The physical layer channel modeler uses the physical layer channel model to infer a plurality of specified positions in the geographic information to obtain a plurality of simulated physical field channel features corresponding to the specified positions. The sparse real physical field channel feature includes a plurality of real physical field channel features (Real Physical Field Channel Feature) respectively measured at a plurality of measurement positions in the geographic information. The entity layer channel training module is connected to the entity channel modeler. The physical layer channel training module receives and utilizes an artificial intelligence algorithm to train geographic information, sparse real physical field channel features, and these simulated physical field channel features to infer a Full real body field channel feature.

In an embodiment of the present invention, the virtual user device signal synthesis system further includes a plurality of transceiver emulator units (Emulator Units). The transceiver simulation units are arranged at the measurement positions in the field to be tested, and are connected to a telecommunication system under test. These transceiver simulation units transmit and receive signals from the telecommunications system under test, while providing sparse real-world field-domain channel features to the entity-layer channel modeler.

In an embodiment of the present invention, the virtual user device signal synthesis system further includes a Geometry Information Fetch Unit. The geographic information extraction unit is connected to the physical layer channel modeler to extract geographic information and provide it to the physical layer channel modeler.

In an embodiment of the present invention, the geographic information acquisition unit is a Lidar, which is used to scan the field to be measured to obtain geographic information.

In an embodiment of the present invention, the virtual user device signal synthesis system further includes: A control unit (Controller) and a virtual user equipment scheduling module (Virtual UE Scheduler). The control unit is connected to the physical layer channel modeler, the physical layer channel training module and the virtual user device scheduling module to control at least one of starting, ending, executing a specified process or step and requesting report data. The control unit configures the number and positions of the plurality of virtual user devices required by the field to be tested. The virtual user device scheduling module performs resource block scheduling, scheduling, management, and message assignment or modification of these virtual user devices.

In an embodiment of the present invention, the physical layer channel training module includes a generator and a discriminator. The generator uses artificial intelligence algorithms to deduce the channel characteristics of the complete real entity field. The classifier uses another artificial intelligence algorithm to judge the authenticity of the complete real entity field channel features generated by the generator, and performs adversarial training between the generator and the classifier until a Nash Equilibrium is reached.

In an embodiment of the present invention, the artificial intelligence algorithm is a Convolution Neural Network (CNN, Convolution Neural Network, -based Algorithm).

The present invention further provides a virtual user device signal synthesis method, which includes the following steps. Receive a geographic information of a field to be measured and a sparse real physical field channel feature to establish a physical layer channel model; use the physical layer channel model to calculate a plurality of designated positions in the geographic information to obtain the corresponding designated positions and receiving and utilizing an artificial intelligence algorithm to train the geographic information, the sparse real-world channel features, and the simulated entity-field channel features to infer that the specified locations and these A complete real-body field channel feature at the measurement location. The above-mentioned sparse real entity field channel features include a plurality of real entity field channel features measured respectively at a plurality of measurement positions in geographic information.

In an embodiment of the present invention, the step of inferring the channel characteristics of the complete real entity field includes the following sub-steps. Use artificial intelligence algorithm to infer the characteristics of the complete real entity field channel; and use another artificial intelligence algorithm to judge the authenticity of the complete real entity field channel feature generated by the generator, and conduct adversarial training between the generator and the classifier, until a Nash equilibrium is reached.

In an embodiment of the present invention, the virtual user device signal synthesis method further includes the following steps. Configure the number and locations of multiple virtual user devices required for the field to be tested. Perform resource block scheduling, scheduling, management, and message assignment or modification of these virtual user devices.

Based on the above, the virtual user device signal synthesis system and method according to the embodiments of the present invention only need to measure the channel characteristics of the real entity at a small number of measurement positions, and then the artificial intelligence algorithm can be used to deduce the complete real entity Field channel feature. Therefore, fast and accurate simulation results can be provided.

The following detailed description will be given in conjunction with the accompanying drawings through specific embodiments, so as to more easily understand the purpose, technical content, characteristics and effects of the present invention.

FIG. 1 is a conceptual diagram illustrating a virtual user device signal synthesis system according to an embodiment of the present invention, and FIG. 2 is a flowchart of a signal synthesis method using the virtual user device signal synthesis system of FIG. 1 . Please refer to FIG. 1 and FIG. 2 together, the virtual user device signal synthesis system 100 includes a physical layer channel modeler 110 and a physical layer channel training module 120 . First, in step S110 , the physical layer channel modeler 110 receives a geographic information G of a field to be measured (not shown) and a sparse real physical field channel feature S to create a physical layer channel model (not shown). ). In this embodiment, the sparse real entity field channel feature S includes a plurality of real entity field channel features D1 measured at a plurality of measurement positions P1 in the geographic information G respectively.

Next, in step S120 , the physical layer channel modeler 110 uses the physical layer channel model to estimate a plurality of designated positions P2 in the geographic information G to obtain a plurality of simulated physical field channel features D2 corresponding to the designated positions P2 . In this embodiment, for example, the physical layer channel model may only consider the geographical relationship between the designated positions P2 and the measurement positions P1, and implement a linear interpolation method to estimate the simulated physical field channel feature D2. Then, in step S130, the physical layer channel training module 120 receives and uses an artificial intelligence algorithm to train the geographic information G, the sparse real physical field channel feature S, and these simulated physical field channel features D2, so as to infer that the A complete real entity field channel feature P of the measurement positions P1 and the specified positions P2. In this embodiment, the artificial intelligence algorithm is, for example, a convolutional neural network road algorithm. The physical layer channel training module 120 includes, but is not limited to, software, hardware, or other algorithms known to assist in machine learning, artificial intelligence, deep learning, neural-like networks, or other equivalent algorithms that can accomplish the same work objectives, Mathematical or human judgment.

It is worth mentioning that this embodiment only needs to measure the real entity field channel feature D1 at a small number of measurement positions P1, and the artificial intelligence algorithm can be used to deduce the complete real entity field channel feature P. Therefore, the present embodiment can provide fast and accurate simulation results. In particular, in the process of training, in addition to considering the geographical relationship between these designated positions P2 and these measurement positions P1, it also considers whether there are obstacles and other environments on these designated positions P2 and these measurement positions P1 condition. Therefore, the inferred complete real entity field channel feature P is more in line with the real situation.

FIG. 3 is a detailed block diagram of the virtual user device signal synthesis system of FIG. 1 . Please refer to FIG. 1 and FIG. 3 , the virtual user device signal synthesis system 100 may further include a geographic information acquisition unit 130 , a plurality of transceiver simulation units 140 , a control unit 150 and a virtual user device scheduling module 160. The geographic information acquisition unit 130 is connected to the physical layer channel modeler 110 for scanning the field to be measured to extract the geographic information G, and provides the physical layer channel modeler 110 . The transceiver analog unit 140 is a hardware device capable of transmitting wireless signals. For example, the transceiver simulation unit 140 can be a Universal Software Radio Peripheral (USRP), an LTE/5G modem with an antenna, or other hardware devices that can achieve the same capability. The geographic information acquisition unit 130 can be a device that provides hardware information, such as LiDAR or other devices that can provide equivalent geographic information. In another embodiment not shown, the virtual user device signal synthesis system 100 may further include a geographic database. The geographic information retrieval unit 130 is connected to the geographic database to obtain geographic information G from the geographic database. That is to say, in addition to using the aforementioned LiDAR or other devices that can provide equivalent geographic information, the geographic information extraction unit 130 can also directly refer to the already stored geographic information G in the geographic database, which can save the need for each time The time required to measure geographic information G is more flexible in use.

The control unit 150 is connected to the physical layer channel modeler 110, the physical layer channel training module 120 and the virtual user device scheduling module 160 to start, stop, execute a specified process or step and request at least one of the reporting data control. That is, the physical layer channel training module 120 can be connected to the physical channel modeler 110 through the control unit 150, but in another illustrated embodiment, the physical layer channel training module 120 can also be connected to the physical channel modeler 110. mold 110. The control unit 150 configures the number and positions of a plurality of virtual user devices (not shown) required by the field to be tested. The virtual user device scheduling module 160 performs resource block scheduling, scheduling, management, and message assignment or modification of these virtual user devices according to the complete real-world feature P.

In this embodiment, the physical layer channel modeler 110 , the physical layer channel training module 120 , the control unit 150 and the virtual terminal device scheduling module 160 include but are not limited to hardware methods such as software or electronic devices and computers. accomplish. The transceiver simulation units 140 are disposed at the measurement positions P1 in the field to be tested, and are connected to a telecommunication system to be tested. The transceiver simulation units 140 transmit and receive signals from the telecommunications system under test, and provide sparse real entity field domain channel features S to the entity layer channel modeler 110 . In this embodiment, the telecommunications system to be tested may be a base station 50 .

Specifically, for a given physical field to be tested, a feature collection phase may be performed first. The user may first place the K transceiver simulation units 140 at the measurement positions P1 in the field to be measured, and then place the base station to be measured 50 at the position to be measured. Next, the transceiver simulation unit 140 will automatically connect to the base station 50 to start communication. Then, according to the 4G\5G standard, the user can obtain a plurality of real physical field channel characteristics D1 according to the downlink transmission frame sent back from the base station 50 to the transceiver simulation unit 140 . In this embodiment, the real physical field channel feature D1 is, for example, channel state information (Channel State Indicator, CSI) or various channel state indicators in 4G\5G specifications).

Considering the sparseness of a small number of transceiver simulation units 140 relative to the physical field to be tested, the user can obtain the channel feature S of the sparse real physical field, and then use the geographic information extraction unit 130 to cooperate with the physical layer channel modeler 110 to obtain the simulated physical field Domain channel feature D2. It is worth noting that the system proposed by the present invention has flexibility, and the user can determine the resolution of the channel characteristics of the field to be measured according to the application requirements, that is, the distance between any two reporting points. After the resolution is determined, using the geographic information extraction unit 130 to cooperate with the physical layer channel modeler 110, the simulated physical field channel feature D2 at the resolution can be obtained. After that, after operating through the AI-assisted simulation stage of the physical layer channel training module 120, the complete real entity field channel feature P at this resolution can be obtained.

FIG. 4 is a flow chart of a signal synthesis method using the virtual user device signal synthesis system of FIG. 3 . Please refer mainly to Figure 4, and refer to Figure 1 and Figure 3 in conjunction. First, in step S210 , the geographic information G is extracted by the geographic information extraction unit 130 and provided to the physical layer channel modeler 110 . While performing step S210, step S240 may also be performed. That is, the transceiver simulation unit 140 provides the sparse real entity field channel feature S to the entity layer channel modeler 110 . In this embodiment, step S240 may be performed periodically, but is not limited thereto.

Next, in step S220 , the physical layer channel modeler 110 receives the geographic information G and the sparse real entity field channel feature S to establish a physical layer channel model. Then in step S230 , the physical layer channel modeler 110 uses the physical layer channel model to estimate the plurality of designated positions P2 in the geographic information G to obtain a plurality of simulated physical field channel features D2 corresponding to the designated positions P4 . Then in step S230 , the physical layer channel modeler 110 uses the physical layer channel model to calculate a plurality of designated positions P2 in the geographic information G to obtain a plurality of simulated physical field channel features D2 corresponding to the designated positions P2 .

Then in step S250, the physical layer channel training module 120 uses an artificial intelligence algorithm to train the geographic information G, the sparse real entity field channel feature S, and these simulated entity field channel features D2 to infer a complete real entity Field channel feature P. In this embodiment, step S250 may be performed periodically, but is not limited thereto.

In addition, while step S240 is performed, step S260 may also be performed, in which the control unit 150 initializes the system and configures the number and positions of multiple virtual user devices (not shown) required in the field to be tested. Then, in step S270, the virtual user equipment scheduling module 160 performs uplink and downlink synchronization (UL/DL) with the telecommunications system under test (base station 50) according to the required number of virtual user equipments. Next, in step S280, the virtual user device scheduling module 160 performs resource block scheduling of these virtual user devices according to the complete real physical field feature P, and receives the message from the control unit 150, and sends the information from the physical layer channel The complete real entity field feature P of the training module 120 is transmitted to the telecommunications system (base station 50) under test. Then, step S290 is performed to determine whether there is an unfinished task. If yes, go back to step S280.

It is worth mentioning that the content of the CSI report reported by the virtual user device and the base station 50 (ie, the complete real physical field feature P) can be transmitted through the simulation unit 140 , the geographic information extraction unit 130 , and the physical layer. The channel modeler 110 and the physical layer channel training module 120 are combined and calculated. In addition, through each training or periodic update, the content of the CSI report of the virtual user device can conform to the framework content of 3GPP 38.214, and other parameters with dependencies can be estimated or assigned, including CQI, Parameter contents such as CRI, PMI, RI, and LI are controlled by the control unit 150 .

In addition, the control unit 150 also performs the uplink/downlink synchronization process defined by the 3GPP 36.211 standard through the transceiver simulation unit 140, including receiving the PSCH signal to determine the Cell ID, comparing it with the SCCH data to achieve time synchronization, checking the PBCH and analyzing the MIB and SIB, and perform subsequent synchronization and setting phases such as PCFICH, PDCCH, PDSCH, and RACH. In this way, the virtual terminal device scheduling module 160 can accept the arrangement of the control unit 150 according to the analyzed data, and transmit the specified message content in the resource block that conforms to the standard. And, finally, the NAS layer completes the establishment of the RRC connection, and communicates with the telecommunications system (base station 50 ) to be tested with subsequent connection establishment and data transmission.

FIG. 5 is a conceptual diagram illustrating a virtual user device signal synthesis system according to another embodiment of the present invention. Please refer to FIG. 1 and FIG. 5 , the virtual user device signal synthesis systems 100 and 200 are similar, the difference is that the physical layer channel training module 220 includes a generator 222 , a classifier 224 and a real database 226 . The generator 222 deduces the complete real entity field channel feature P by using an artificial intelligence algorithm. The classifier 224 uses another artificial intelligence algorithm to judge the authenticity of the complete real entity field channel feature P generated by the generator 222, and performs adversarial training between the generator 222 and the classifier 224 until a Nash equilibrium is reached. The real data base 226 provides real data for the classifier 224 to make training judgments.

That is to say, this embodiment is described with an adversarial generative network architecture (Generative Adversarial Network, GAN) for assistance, but the AI architecture applied in the present invention includes but is not limited to the GAN network architecture. The working objective of the generator 222 is to infer the complete real entity field channel feature P according to the sparse real entity field channel feature S. To help the generator 222 achieve this goal, another AI algorithm will be used to train the classifier 224 in the future to judge the authenticity of the complete real-world channel features generated by the generator 222 . After the confrontation training between the generator 222 and the classifier 224, when the Nash equilibrium is reached, the generator 222 will have the ability to generate highly realistic and complete real entity field channel features, that is, the AI-assisted simulation stage training process is completed.

It is worth noting that there is also flexibility in the generation of the ground truth dataset (Ground truth Dataset) of the ground truth database 226 . In this embodiment, the complete real entity field channel feature P or the simulated entity field channel feature P2 can be post-processed to obtain the data set and regard it as the label corresponding to the sample. In conclusion, the proposed AI-assisted simulation stage model training belongs to a kind of supervised learning, and the main purpose is to train the generator 222 to simulate a corresponding complete sample according to a sparse and incomplete sample. The purpose of designing the classifier 224 is to provide a special loss function to guide the generator 222 to generate the complete real entity field channel feature P that corresponds to the actual situation.

After the training process is completed, this embodiment only needs to give a sample of a new scene (including sparse real physical field channel features, simulated physical field channel features and physical field geographic information) and feed it into the physical layer channel training module. , the entity layer channel training module can infer the complete real entity field channel characteristics, and provide the channel characteristic parameters of the specified geographic location to other system components. It is worth noting that in the testing stage, it is not necessary to provide the complete real entity field channel characteristics. The trained AI model can infer its measurement results based on the incomplete samples, so that the complete samples can be greatly reduced. Measure the required manpower and material resources to achieve the design purpose of this embodiment.

To sum up, since the present invention introduces an AI algorithm to realize the physical layer channel training module, and bridges the two methods described in the prior art. Therefore, the present invention can use the sparse real field channel features measured by a small number of transceiver simulation units and the simulated entity field channel characteristics obtained by channel modeling, so that the physical layer channel training module can infer the real quantity of a large number of transceiver simulation units. The results of the measured channel characteristics of the complete real entity field. In addition, the user can obtain the field channel characteristic measurement result in a very short time through the virtual user device signal synthesis method proposed by the present invention. Moreover, with the increase of time and training times, the accuracy can be continuously improved. In addition, after the virtual terminal device scheduling module receives the field-domain channel feature measurement results, the upper-layer parameter control of the control unit and the behavior design of the user device, and the time domain/frequency domain of the virtual terminal device scheduling module The resource block scheduling can send the signal of the virtual user device, so that the telecommunications system under test can be identified as the signal of the virtual user device of the specified geographical location. In this way, a fixed number of transceiving analog units are finally realized to send signals of virtual terminal devices in each geographic location to complete the field test to be tested.

50: Base Station 100, 200: Virtual user device signal synthesis system 110: Entity Layer Channel Modeler 120: Entity layer channel training module 130: Geographic Information Retrieval Unit 140: Transceiver analog unit 150: Control unit 160:Virtual User Device Scheduling Module 222: Generator 224: Classifier 226: Real Database G: geographic information D1: Real entity field channel feature D2: Simulate Solid Field Channel Features P: full real entity field channel feature P1: measuring position P2: Specify the location S: Sparse real entity field channel feature S110~S130, S210~S290: Steps

FIG. 1 is a conceptual diagram illustrating a virtual user device signal synthesis system according to an embodiment of the present invention. FIG. 2 is a flow chart of a signal synthesis method using the virtual user device signal synthesis system of FIG. 1 . FIG. 3 is a detailed block diagram of the virtual user device signal synthesis system of FIG. 1 . FIG. 4 is a flow chart of a signal synthesis method using the virtual user device signal synthesis system of FIG. 3 . FIG. 5 is a conceptual diagram illustrating a virtual user device signal synthesis system according to another embodiment of the present invention.

100: Virtual User Device Signal Synthesis System

110: Entity Layer Channel Modeler

120: Entity layer channel training module

G: geographic information

D1: Real entity field channel feature

D2: Simulate Solid Field Channel Features

P1: measuring position

P2: Specify the location

S: Sparse real entity field channel feature

Claims (11)

A virtual user device signal synthesis system, comprising: A physical layer channel modeler receives a geographic information of a field to be measured and a sparse real physical field channel feature to build a physical layer channel model, and uses the physical layer channel model Performing estimation at specified positions to obtain a plurality of simulated entity field channel features corresponding to the specified positions, wherein the sparse real entity field channel features include a plurality of respectively measured at a plurality of measurement positions in the geographic information Real Solid Field Channel Features; and A physical layer channel training module, connected to the physical channel modeler, receives and uses an artificial intelligence algorithm to train the geographic information, the sparse real physical field channel features and the simulated physical field channel features, to infer a complete real entity field channel feature covering the specified locations and the measured locations. The virtual user device signal synthesis system as claimed in claim 1, further comprising: A plurality of transceiver simulation units are arranged at the measurement positions in the field to be tested and connected to a telecommunications system to be tested, wherein the transceiver simulation units transmit and receive signals to the telecommunications system to be tested to provide the sparse reality The solid field channel feature to the solid layer channel modeler. The virtual user device signal synthesis system as claimed in claim 1, further comprising: A geographic information extraction unit is connected to the physical layer channel modeler to extract the geographic information and provide the physical layer channel modeler. The virtual user device signal synthesis system as claimed in claim 3, wherein the geographic information acquisition unit is a lidar, used for scanning the to-be-measured field to obtain the geographic information. The virtual user device signal synthesis system according to claim 3, further comprising: A geographic database, wherein the geographic information retrieval unit is connected to the geographic database to obtain the geographic information from the geographic database. The virtual user device signal synthesis system as claimed in claim 1, further comprising: a control unit, connected to the physical layer channel modeler, the physical layer channel training module and the virtual user device scheduling module, for starting, ending, executing a specified process or step, and requesting at least one of the reporting data a control, and configure the number and location of a plurality of virtual user devices required for the field to be tested; and A virtual user device scheduling module performs resource block scheduling, scheduling, management, and message assignment or modification of the virtual user devices according to the characteristics of the complete real entity field. The virtual user device signal synthesis system as claimed in claim 1, wherein the physical layer channel training module comprises: a generator, using the artificial intelligence algorithm to deduce the channel characteristics of the complete real entity field; and a classifier, which uses another artificial intelligence algorithm to judge the authenticity of the complete real entity field channel features generated by the generator, and performs adversarial training between the generator and the classifier until a Naxus equilibrium is reached. The virtual user device signal synthesis system according to claim 1, wherein the artificial intelligence algorithm is a convolutional neural network road algorithm. A method for synthesizing a virtual user device signal, comprising: Receive a geographic information of a field to be measured and a sparse real entity field channel feature to establish a physical layer channel model, wherein the sparse real entity field channel feature includes all measurement positions in the geographic information. Multiple real entity field channel features measured separately; Using the physical layer channel model to infer a plurality of specified locations in the geographic information to obtain a plurality of simulated entity field channel features corresponding to the specified locations; and receiving and using an artificial intelligence algorithm to train the geographic information, the sparse real physical field channel features, and the simulated physical field channel features to infer a complete set of the specified locations and the measured locations Real solid field channel feature. The virtual user device signal synthesis method as claimed in claim 9, wherein the step of inferring the characteristics of the complete real entity field channel comprises: Inferring the complete real entity field channel characteristics using the artificial intelligence algorithm; and Use another artificial intelligence algorithm to judge the authenticity of the complete real entity field channel features generated by the generator, and conduct adversarial training between the generator and the classifier until a Nash Equilibrium is reached. The virtual user device signal synthesis method as claimed in claim 9, further comprising: the number and location of the plurality of virtual user devices required to configure the field under test; and According to the characteristics of the complete real physical field, the resource block scheduling, scheduling, management, and message assignment or modification of the virtual user devices are performed.

Images